[quote] In general having fewer faster processors is better because there are times when you're going to be waiting on a single thread. <hr></blockquote>
So that means a dual is better then a quad because it has fewr processors? So if hypothetically a dual 1.2ghz and a quad 867 come out the dual 1.2ghz would be faster or better? Then what's the point of a quad?
<strong>So that means a dual is better then a quad because it has fewr processors? So if hypothetically a dual 1.2ghz and a quad 867 come out the dual 1.2ghz would be faster or better? Then what's the point of a quad?</strong><hr></blockquote>
No, no, no... it also depends very much on how heavily threaded your apps are...
If you have an old app with zero multithreading, that single thread can only run on one processor. There is no way to "split" a thread to work on multiple processors. If you have a quad 800 MHz Mac, a single-threaded app would only be able to use a single 800 MHz chip. In this case, a single 867 MHz CPU should be better.
However, newer apps are are becoming much more threaded because of Mac OS X's oo-nature. If you have a very multi-threaded app, it'll run great with many many processors. If said app had a hundred threads, it'd be best run on a hundred processors, even if they're much slower processors. In this case, a quad 350 MHz Mac would (*in theory*) run circles around a single 867 MHz Mac because all the threads can be evenly distributed to use up all of the four slower processors (which -again *in theory*- actually amount to 1400 MHz!). Of course, that doesn't mean it won't run well on single processor machines; the OS will simply put all the threads on the single CPU (which would have to be very fast to evenly compete with the multi-procs).
Right now Apple has chosen the best of these two paths. They are selling the mid-speed CPU as a duallie. Unless Motorola starts getting huge gains in their yields, we won't be seeing multiple processors across the board (although that would be sweet!).
Am I making any sense yet? <img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
If you have an old app with zero multithreading, that single thread can only run on one processor. There is no way to "split" a thread to work on multiple processors. If you have a quad 800 MHz Mac, a single-threaded app would only be able to use a single 800 MHz chip. In this case, a single 867 MHz CPU should be better.
Am I making any sense yet? <img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" /> </strong><hr></blockquote>
No you are not, if you have 2 CPUs you can use multi-threading, where your 1 thread application may occupy 1 CPU, but your OS can still use the other CPU to do other work. If you have 4 CPUs you can have 3 1 thread apps running at about 90% and still be able to do other work.
There is no scenario where 1 CPU is better than 2, unless you want something done slower rather than faster. Or if your OS does not support multithreading at all.
No you are not, if you have 2 CPUs you can use multi-threading, where your 1 thread application may occupy 1 CPU, but your OS can still use the other CPU to do other work. If you have 4 CPUs you can have 3 1 thread apps running at about 90% and still be able to do other work.
There is no scenario where 1 CPU is better than 2, unless you want something done slower rather than faster. Or if your OS does not support multithreading at all.</strong><hr></blockquote>
Couldn't you tell I was using exagerrated examples above? I don't think you were paying close attention to my examples above or perhaps I need to clarify...
Yes, there are scenarios where 1 CPU is better that 2. I gave you one above with the dual 800 versus single 867. If you have looked at your Process Viewer or 'top' lately, you'll see that the OS itself is not using much of the processor. In theory (I have no single-thread apps like this nor several Macs to test on), a CPU-hungry app with a single thread will take over one of the 800 MHz procs in the dual setup while the OS uses the other. If I am running only that app, the other processor sits idle, using maybe 2 or 3 percent for system tasks. However, on the single 867, the OS and app have to share the same proc, but because the OS's use is so small, the majority of that 67 MHz difference goes to that CPU-hungry thread, thus making the single 867 beat out the dual 800.
Of course, this is not a common example because most people today know to multitask between several running apps. Yet, it is a valid, plausible example.
<strong>Yes, there are scenarios where 1 CPU is better that 2. I gave you one above with the dual 800 versus single 867. If you have looked at your Process Viewer or 'top' lately, you'll see that the OS itself is not using much of the processor. In theory (I have no single-thread apps like this nor several Macs to test on), a CPU-hungry app with a single thread will take over one of the 800 MHz procs in the dual setup while the OS uses the other. If I am running only that app, the other processor sits idle, using maybe 2 or 3 percent for system tasks. However, on the single 867, the OS and app have to share the same proc, but because the OS's use is so small, the majority of that 67 MHz difference goes to that CPU-hungry thread, thus making the single 867 beat out the dual 800.</strong><hr></blockquote>
I still think the dual 800 is faster, simply because the OS overhead can be dealt with much easier with the dual system - in a reasonable scenarion (if the OS is well coded) you can work with 100% of one 800 CPU on that one thread, without the need to switch threads to the OS itself, which means a hell of a less flushing and reloading data in and out of the caches, ram and the CPU registers. It's much better on the G4 than the P4 because of the shorter CPU pipeline, however.
If there are people who only run 1 app at a time, then a single CPU system is enough, and you are right, but in most cases people tend to open iTunes or Internet Explorer to entertain themselves where the real strenght of a dual system shows.
If we take into consideration that a lot of people still run OS9 a single CPU system is as well enough, since only OSX can get full advantage of a dual system.
I still think the dual 800 is faster, simply because the OS overhead can be dealt with much easier with the dual system - in a reasonable scenarion (if the OS is well coded) you can work with 100% of one 800 CPU on that one thread, without the need to switch threads to the OS itself, which means a hell of a less flushing and reloading data in and out of the caches, ram and the CPU registers. It's much better on the G4 than the P4 because of the shorter CPU pipeline, however.
If there are people who only run 1 app at a time, then a single CPU system is enough, and you are right, but in most cases people tend to open iTunes or Internet Explorer to entertain themselves where the real strenght of a dual system shows.
If we take into consideration that a lot of people still run OS9 a single CPU system is as well enough, since only OSX can get full advantage of a dual system.</strong><hr></blockquote>
A couple of points.
Firstly, Apple uses a shared bus architecture for their multiprocessor machines, that means that all the processors share the bandwidth of a single (currently very slow) bus. Even a multithreaded app. running on 4 processors may well suffer because of this, there are very few circumstances where a programme is so heavily compute bound that it would see a noticeable speed up under these circumstances, all the processors would be continually fighting for bandwidth. I suspect this is why Apple have'nt released a 4 processor machine, they need to change the bus structure first.
Secondly, personally I run a lot of scientific modelling software, some of which cannot really be multithreaded (I know, I write it) by it's nature, and I want that to run as fast as possible. Adding a second processor would actually slow down this work because of the overhead involved in managing multiple processors.
How could they fit four of the chips into their enclosure, and no,
putting two different chips into their pro machine would not be good, because then people would be hesitant of which one to get, not as simple as you thought m'lady!
Firstly, Apple uses a shared bus architecture for their multiprocessor machines ...
Secondly, personally I run a lot of scientific modelling software, some of which cannot really be multithreaded (I know, I write it) by it's nature, and I want that to run as fast as possible. Adding a second processor would actually slow down this work because of the overhead involved in managing multiple processors.
Michael</strong><hr></blockquote>
I didn't know they use a shared bus, but at the speed of Apples bus it really does not make much sense then, yes.
As for multithreading - from as far as I know, it mostly depends on the multi-CPU architecture you are working with. I was assuming the dual/multi CPU environment as each CPU having it's own "bus", cache and ram bank. Also I think that a lot of scientific software can be actually breaked down into separate threads (depends on the software architect and the tools at his disposal also) which would increase performance on a multi-CPU machine, it's more a matter of what solution one needs.
There is no scenario where 1 CPU is better than 2, unless you want something done slower rather than faster. Or if your OS does not support multithreading at all.</strong><hr></blockquote>
This is patently false... if you are running a long computation series which is completely dependent on the previous result then you are restricted to the speed of a single processor. Due to practical constraints if you have a single processor machine its single-threaded performance will most likely be higher than if you had a dual processor machine (for many reasons: fixed budget, shared hardware resources, power consumption, heat dissipation, etc).
Some tasks just don't thread well, or the software's author hasn't spent the time to thread it. Either way having a multi-processor isn't going to help you. If you spend most of your time waiting for a specific task to finish and you are doing nothing else then you would have been better off to buy a single processor at a higher clock rate than to buy a dual processor at a lesser clock rate.
If all the processors in all available machines are of an equal clock rate (I'm talking about processors of uniform type, by the way), then yes a dual processor machine will be the faster one. If you're bound waiting for a single task then the second processor will generally sit idle, not consuming any bus or memory cycles.
This argument extends to quad processor machines in the same way... if you have a fixed amount of money then you can either have 4 processors at a slower speed, 2 at a higher speed, or 1 at the highest speed. If you are multi-threaded (like MacOS X) then it makes sense to go multi-processor if you can't possibly get a faster single processor and still have money to spend. If you are multi-threaded and have the fastest possible dual processor, then going to a quad processor will be an improvement.
Think of this: if you could have 1000 1 MHz processors or 1 1000 MHz processor, and you had to wait for one computation to finish on one processor and it takes 1000 MHz/hours of computation... would you rather wait for 1 hour or a thousand?
Conversely, if you had 2 computations that take 1000 MHz/hours each and you could have a single 1000 MHz processor, or a dual 1000 MHz processor... would you rather wait 2 hours or 1 hour?
[In this last example I ignored performance losses due to bus contention and the like, depending on the MP hardware, algorithms, data set sizes and cache sizes. A multiprocessor system usually loses 10-25% of its theoretical performance to these factors. The processors essentially get in eachother's way, and have to spend time coordinating their activities]
[One other note: not all threads run all the time. Part of the advantage of threaded programming is that threads can go to "sleep" waiting for some event to happen, like the user clicking or data arriving from the network. In these cases giving sleeping threads their own processor saves you nothing because those threads aren't doing any computation that you're waiting for]
If G4's come out and it's say a dual 1.2 as the top-what would the specs need to be (bus DDR) to make it a very worthy purchase?
ie. the bus staying as current wouldn't seem to do the duals justice. So what would the bus need to be? Also, I assume that DDR will also make a huge difference. I assume that DDR usage would likely mean a faster bus? Is that a given?
Basically, if my fears come to life and there is no G5, what would be acceptable. I don't think faster G4s in current mobo architecture would be acceptable. Right?
If G4's come out and it's say a dual 1.2 as the top-what would the specs need to be (bus DDR) to make it a very worthy purchase?<hr></blockquote>The simple answer is that there is no simple answer! There are too many possible variables that we just don't know about to make any conclusions. You'd have to seriously consider the amounts of L1, L2, and L3 cache, the bus speed, the RAM speed, the (probably somewhat different) architechture or the chip, *price*, etc. etc. I don't like making suggestions about products that I have *zero* real info about.
[quote]Basically, if my fears come to life and there is no G5, what would be acceptable. I don't think faster G4s in current mobo architecture would be acceptable. Right?[/qb]<hr></blockquote>Well, some people argue that in current Macs the bus is already a bottleneck. But who is to say that Apple won't put faster G4's in new mobos?
<strong>Basically, if my fears come to life and there is no G5, what would be acceptable. I don't think faster G4s in current mobo architecture would be acceptable. Right?</strong><hr></blockquote>
Heh, only you can set your own definition of acceptable. Acceptable for what? One thing that gets lost in all the discussions and machine comparisons is that they are all damn fast machines. For most applications the current top-of-the-line computers are more than adequate. People that use compute-bound applications will obviously always want faster machines (at least until their task happens instantly, or in real-time depending on the nature of it), but how many people does that really cover? What is more interesting are killer-apps enabled by a sudden jump in performance. What can the new machine do for me that previous machines could not?
Heh, only you can set your own definition of acceptable. Acceptable for what? One thing that gets lost in all the discussions and machine comparisons is that they are all damn fast machines. For most applications the current top-of-the-line computers are more than adequate. People that use compute-bound applications will obviously always want faster machines (at least until their task happens instantly, or in real-time depending on the nature of it), but how many people does that really cover? What is more interesting are killer-apps enabled by a sudden jump in performance. What can the new machine do for me that previous machines could not?</strong><hr></blockquote>
Well, for me it comes down to 2 things
1) why drop $5k (22" monitor) on same G4 but over 1ghz now when next gen G5 will be out in 6 months?
2) I want my next purchase to last as long as it can. I have a G4 450 and it still does what it needs, but it's started to jog instead of sprint I know Photoshop 8 will probably need 8oomegs of ram and I want to get the best machine for that with my money.
That's why I want to know if no G5s what's the most realistic, logical rev that I should consider? I don't think a speed bump to the current QS is good enough. Not when we have DDR, faster buses, and RIO and all that good tech stuff right around the corner. Unless they come out with a new G4, but I just want some good tech advice. What to look for, what is acceptable. I guess our current bus isn't acceptable.
1) why drop $5k (22" monitor) on same G4 but over 1ghz now when next gen G5 will be out in 6 months?
2) I want my next purchase to last as long as it can. I have a G4 450 and it still does what it needs, but it's started to jog instead of sprint I know Photoshop 8 will probably need 8oomegs of ram and I want to get the best machine for that with my money.
That's why I want to know if no G5s what's the most realistic, logical rev that I should consider? I don't think a speed bump to the current QS is good enough. Not when we have DDR, faster buses, and RIO and all that good tech stuff right around the corner. Unless they come out with a new G4, but I just want some good tech advice. What to look for, what is acceptable. I guess our current bus isn't acceptable.</strong><hr></blockquote>
1) There is always something better just around the corner in the computer business. If you don't need a new machine for something specific then by waiting you can always get something better, and probably for cheaper. Eventually something will come along that is "must have".
2) That G4 will last just fine until some new capability comes along that you need, or some new killer-app arrives that has to run in real-time and thus requires a certain level of processing power. Without hard performance requirements, speed isn't a compelling reason to upgrade.
Comments
So that means a dual is better then a quad because it has fewr processors? So if hypothetically a dual 1.2ghz and a quad 867 come out the dual 1.2ghz would be faster or better? Then what's the point of a quad?
<strong>So that means a dual is better then a quad because it has fewr processors? So if hypothetically a dual 1.2ghz and a quad 867 come out the dual 1.2ghz would be faster or better? Then what's the point of a quad?</strong><hr></blockquote>
No, no, no... it also depends very much on how heavily threaded your apps are...
If you have an old app with zero multithreading, that single thread can only run on one processor. There is no way to "split" a thread to work on multiple processors. If you have a quad 800 MHz Mac, a single-threaded app would only be able to use a single 800 MHz chip. In this case, a single 867 MHz CPU should be better.
However, newer apps are are becoming much more threaded because of Mac OS X's oo-nature. If you have a very multi-threaded app, it'll run great with many many processors. If said app had a hundred threads, it'd be best run on a hundred processors, even if they're much slower processors. In this case, a quad 350 MHz Mac would (*in theory*) run circles around a single 867 MHz Mac because all the threads can be evenly distributed to use up all of the four slower processors (which -again *in theory*- actually amount to 1400 MHz!). Of course, that doesn't mean it won't run well on single processor machines; the OS will simply put all the threads on the single CPU (which would have to be very fast to evenly compete with the multi-procs).
Right now Apple has chosen the best of these two paths. They are selling the mid-speed CPU as a duallie. Unless Motorola starts getting huge gains in their yields, we won't be seeing multiple processors across the board (although that would be sweet!).
Am I making any sense yet? <img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
[ 12-30-2001: Message edited by: starfleetX ]</p>
<strong>
If you have an old app with zero multithreading, that single thread can only run on one processor. There is no way to "split" a thread to work on multiple processors. If you have a quad 800 MHz Mac, a single-threaded app would only be able to use a single 800 MHz chip. In this case, a single 867 MHz CPU should be better.
Am I making any sense yet? <img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
No you are not, if you have 2 CPUs you can use multi-threading, where your 1 thread application may occupy 1 CPU, but your OS can still use the other CPU to do other work. If you have 4 CPUs you can have 3 1 thread apps running at about 90% and still be able to do other work.
There is no scenario where 1 CPU is better than 2, unless you want something done slower rather than faster. Or if your OS does not support multithreading at all.
<strong>
No you are not, if you have 2 CPUs you can use multi-threading, where your 1 thread application may occupy 1 CPU, but your OS can still use the other CPU to do other work. If you have 4 CPUs you can have 3 1 thread apps running at about 90% and still be able to do other work.
There is no scenario where 1 CPU is better than 2, unless you want something done slower rather than faster. Or if your OS does not support multithreading at all.</strong><hr></blockquote>
Couldn't you tell I was using exagerrated examples above?
Yes, there are scenarios where 1 CPU is better that 2. I gave you one above with the dual 800 versus single 867. If you have looked at your Process Viewer or 'top' lately, you'll see that the OS itself is not using much of the processor. In theory (I have no single-thread apps like this nor several Macs to test on), a CPU-hungry app with a single thread will take over one of the 800 MHz procs in the dual setup while the OS uses the other. If I am running only that app, the other processor sits idle, using maybe 2 or 3 percent for system tasks. However, on the single 867, the OS and app have to share the same proc, but because the OS's use is so small, the majority of that 67 MHz difference goes to that CPU-hungry thread, thus making the single 867 beat out the dual 800.
Of course, this is not a common example because most people today know to multitask between several running apps. Yet, it is a valid, plausible example.
<strong>Yes, there are scenarios where 1 CPU is better that 2. I gave you one above with the dual 800 versus single 867. If you have looked at your Process Viewer or 'top' lately, you'll see that the OS itself is not using much of the processor. In theory (I have no single-thread apps like this nor several Macs to test on), a CPU-hungry app with a single thread will take over one of the 800 MHz procs in the dual setup while the OS uses the other. If I am running only that app, the other processor sits idle, using maybe 2 or 3 percent for system tasks. However, on the single 867, the OS and app have to share the same proc, but because the OS's use is so small, the majority of that 67 MHz difference goes to that CPU-hungry thread, thus making the single 867 beat out the dual 800.</strong><hr></blockquote>
I still think the dual 800 is faster, simply because the OS overhead can be dealt with much easier with the dual system - in a reasonable scenarion (if the OS is well coded) you can work with 100% of one 800 CPU on that one thread, without the need to switch threads to the OS itself, which means a hell of a less flushing and reloading data in and out of the caches, ram and the CPU registers. It's much better on the G4 than the P4 because of the shorter CPU pipeline, however.
If there are people who only run 1 app at a time, then a single CPU system is enough, and you are right, but in most cases people tend to open iTunes or Internet Explorer to entertain themselves where the real strenght of a dual system shows.
If we take into consideration that a lot of people still run OS9 a single CPU system is as well enough, since only OSX can get full advantage of a dual system.
<strong>
I still think the dual 800 is faster, simply because the OS overhead can be dealt with much easier with the dual system - in a reasonable scenarion (if the OS is well coded) you can work with 100% of one 800 CPU on that one thread, without the need to switch threads to the OS itself, which means a hell of a less flushing and reloading data in and out of the caches, ram and the CPU registers. It's much better on the G4 than the P4 because of the shorter CPU pipeline, however.
If there are people who only run 1 app at a time, then a single CPU system is enough, and you are right, but in most cases people tend to open iTunes or Internet Explorer to entertain themselves where the real strenght of a dual system shows.
If we take into consideration that a lot of people still run OS9 a single CPU system is as well enough, since only OSX can get full advantage of a dual system.</strong><hr></blockquote>
A couple of points.
Firstly, Apple uses a shared bus architecture for their multiprocessor machines, that means that all the processors share the bandwidth of a single (currently very slow) bus. Even a multithreaded app. running on 4 processors may well suffer because of this, there are very few circumstances where a programme is so heavily compute bound that it would see a noticeable speed up under these circumstances, all the processors would be continually fighting for bandwidth. I suspect this is why Apple have'nt released a 4 processor machine, they need to change the bus structure first.
Secondly, personally I run a lot of scientific modelling software, some of which cannot really be multithreaded (I know, I write it) by it's nature, and I want that to run as fast as possible. Adding a second processor would actually slow down this work because of the overhead involved in managing multiple processors.
Michael
putting two different chips into their pro machine would not be good, because then people would be hesitant of which one to get, not as simple as you thought m'lady!
<strong>
Firstly, Apple uses a shared bus architecture for their multiprocessor machines ...
Secondly, personally I run a lot of scientific modelling software, some of which cannot really be multithreaded (I know, I write it) by it's nature, and I want that to run as fast as possible. Adding a second processor would actually slow down this work because of the overhead involved in managing multiple processors.
Michael</strong><hr></blockquote>
I didn't know they use a shared bus, but at the speed of Apples bus it really does not make much sense then, yes.
As for multithreading - from as far as I know, it mostly depends on the multi-CPU architecture you are working with. I was assuming the dual/multi CPU environment as each CPU having it's own "bus", cache and ram bank. Also I think that a lot of scientific software can be actually breaked down into separate threads (depends on the software architect and the tools at his disposal also) which would increase performance on a multi-CPU machine, it's more a matter of what solution one needs.
<strong>
There is no scenario where 1 CPU is better than 2, unless you want something done slower rather than faster. Or if your OS does not support multithreading at all.</strong><hr></blockquote>
This is patently false... if you are running a long computation series which is completely dependent on the previous result then you are restricted to the speed of a single processor. Due to practical constraints if you have a single processor machine its single-threaded performance will most likely be higher than if you had a dual processor machine (for many reasons: fixed budget, shared hardware resources, power consumption, heat dissipation, etc).
Some tasks just don't thread well, or the software's author hasn't spent the time to thread it. Either way having a multi-processor isn't going to help you. If you spend most of your time waiting for a specific task to finish and you are doing nothing else then you would have been better off to buy a single processor at a higher clock rate than to buy a dual processor at a lesser clock rate.
If all the processors in all available machines are of an equal clock rate (I'm talking about processors of uniform type, by the way), then yes a dual processor machine will be the faster one. If you're bound waiting for a single task then the second processor will generally sit idle, not consuming any bus or memory cycles.
This argument extends to quad processor machines in the same way... if you have a fixed amount of money then you can either have 4 processors at a slower speed, 2 at a higher speed, or 1 at the highest speed. If you are multi-threaded (like MacOS X) then it makes sense to go multi-processor if you can't possibly get a faster single processor and still have money to spend. If you are multi-threaded and have the fastest possible dual processor, then going to a quad processor will be an improvement.
Think of this: if you could have 1000 1 MHz processors or 1 1000 MHz processor, and you had to wait for one computation to finish on one processor and it takes 1000 MHz/hours of computation... would you rather wait for 1 hour or a thousand?
Conversely, if you had 2 computations that take 1000 MHz/hours each and you could have a single 1000 MHz processor, or a dual 1000 MHz processor... would you rather wait 2 hours or 1 hour?
[In this last example I ignored performance losses due to bus contention and the like, depending on the MP hardware, algorithms, data set sizes and cache sizes. A multiprocessor system usually loses 10-25% of its theoretical performance to these factors. The processors essentially get in eachother's way, and have to spend time coordinating their activities]
[One other note: not all threads run all the time. Part of the advantage of threaded programming is that threads can go to "sleep" waiting for some event to happen, like the user clicking or data arriving from the network. In these cases giving sleeping threads their own processor saves you nothing because those threads aren't doing any computation that you're waiting for]
[ 12-30-2001: Message edited by: Programmer ]</p>
If G4's come out and it's say a dual 1.2 as the top-what would the specs need to be (bus DDR) to make it a very worthy purchase?
ie. the bus staying as current wouldn't seem to do the duals justice. So what would the bus need to be? Also, I assume that DDR will also make a huge difference. I assume that DDR usage would likely mean a faster bus? Is that a given?
Basically, if my fears come to life and there is no G5, what would be acceptable. I don't think faster G4s in current mobo architecture would be acceptable. Right?
If G4's come out and it's say a dual 1.2 as the top-what would the specs need to be (bus DDR) to make it a very worthy purchase?<hr></blockquote>The simple answer is that there is no simple answer!
[quote]Basically, if my fears come to life and there is no G5, what would be acceptable. I don't think faster G4s in current mobo architecture would be acceptable. Right?[/qb]<hr></blockquote>Well, some people argue that in current Macs the bus is already a bottleneck. But who is to say that Apple won't put faster G4's in new mobos?
[edit] ugly UBB code
[ 12-31-2001: Message edited by: starfleetX ]</p>
<strong>Basically, if my fears come to life and there is no G5, what would be acceptable. I don't think faster G4s in current mobo architecture would be acceptable. Right?</strong><hr></blockquote>
Heh, only you can set your own definition of acceptable. Acceptable for what? One thing that gets lost in all the discussions and machine comparisons is that they are all damn fast machines. For most applications the current top-of-the-line computers are more than adequate. People that use compute-bound applications will obviously always want faster machines (at least until their task happens instantly, or in real-time depending on the nature of it), but how many people does that really cover? What is more interesting are killer-apps enabled by a sudden jump in performance. What can the new machine do for me that previous machines could not?
<strong>
Heh, only you can set your own definition of acceptable. Acceptable for what? One thing that gets lost in all the discussions and machine comparisons is that they are all damn fast machines. For most applications the current top-of-the-line computers are more than adequate. People that use compute-bound applications will obviously always want faster machines (at least until their task happens instantly, or in real-time depending on the nature of it), but how many people does that really cover? What is more interesting are killer-apps enabled by a sudden jump in performance. What can the new machine do for me that previous machines could not?</strong><hr></blockquote>
Well, for me it comes down to 2 things
1) why drop $5k (22" monitor) on same G4 but over 1ghz now when next gen G5 will be out in 6 months?
2) I want my next purchase to last as long as it can. I have a G4 450 and it still does what it needs, but it's started to jog instead of sprint
That's why I want to know if no G5s what's the most realistic, logical rev that I should consider? I don't think a speed bump to the current QS is good enough. Not when we have DDR, faster buses, and RIO and all that good tech stuff right around the corner. Unless they come out with a new G4, but I just want some good tech advice. What to look for, what is acceptable. I guess our current bus isn't acceptable.
<strong>
Well, for me it comes down to 2 things
1) why drop $5k (22" monitor) on same G4 but over 1ghz now when next gen G5 will be out in 6 months?
2) I want my next purchase to last as long as it can. I have a G4 450 and it still does what it needs, but it's started to jog instead of sprint
That's why I want to know if no G5s what's the most realistic, logical rev that I should consider? I don't think a speed bump to the current QS is good enough. Not when we have DDR, faster buses, and RIO and all that good tech stuff right around the corner. Unless they come out with a new G4, but I just want some good tech advice. What to look for, what is acceptable. I guess our current bus isn't acceptable.</strong><hr></blockquote>
1) There is always something better just around the corner in the computer business. If you don't need a new machine for something specific then by waiting you can always get something better, and probably for cheaper. Eventually something will come along that is "must have".
2) That G4 will last just fine until some new capability comes along that you need, or some new killer-app arrives that has to run in real-time and thus requires a certain level of processing power. Without hard performance requirements, speed isn't a compelling reason to upgrade.